A Novel Method for Iris Recognition Using Fusion of Wavelets and DFT

Abstract

A robust approach for iris recognition using wavelet based feature extraction and decision level fusion is proposed. In this method, circular Hough transform is used for iris segmentation and Daugman's rubber sheet model for normalization. For feature extraction, a combination of Haar wavelet decomposition and spectral transformation of 1D log Gabor wavelet transform is used. Discrete Fourier transform (DFT) is used as spectral transformation tool. The spectral transformation reduces the redundancy of the feature vectors, which adds the recognition rate. Euclidean distance classifier is used for classification and decision level fusion is employed. The experimental results shows that the proposed method gives better performance. CASIA database is used for evaluation. Index Terms-Iris recognition, Haar wavelet, 1D log gabor wavelet, Euclidean distance, decision level fusion.